GEMM RCCL Overlap Comprehensive Performance Report
Complete Analysis: rocm-7.0.8-meta (Baseline) vs rocm-7.0.10-meta (Test)
Executive Summary
This comprehensive report contains all performance analysis plots and metrics for RCCL comparisons across multiple configurations.
Test Configuration
- Baseline: rocm-7.0.8-meta
- Test Version: rocm-7.0.10-meta
- Configurations: 8 total (256/512 threads × 28/42/56/70 channels)
- Total Plots: 96 visualizations
Configuration: 256 Threads, 28 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 256 Threads, 42 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 256 Threads, 56 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 256 Threads, 70 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 512 Threads, 28 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 512 Threads, 42 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 512 Threads, 56 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
Configuration: 512 Threads, 70 Channels
Base: rocm-7.0.8-meta
Test: rocm-7.0.10-meta
Test: rocm-7.0.10-meta
Overview Plots
Percentage Change Overview
Absolute Time Comparison
Performance Heatmap
Total Execution Time by Rank
Detailed Metrics
Computation Time Across Ranks
Communication Time Across Ranks
Idle Time Across Ranks
Percentage Difference All Metrics
NCCL Analysis
NCCL Latency Analysis
NCCL Summary Analysis
